On Tuesday, Advanced Micro Devices Inc (AMD.O) revealed new details about an artificial intelligence processor that will challenge market leader Nvidia Corp (NVDA.O). Still, Wall Street needed to know who would buy it.
Santa Clara-based AMD claimed the chip, which would start trickling out in the third quarter and mass production in the fourth, will have 192 gigabytes of memory.
In an interview with Reuters, AMD CEO Lisa Su said it might help tech companies control the rising cost of ChatGPT-like services. Su gave a speech in San Francisco in which an AI system using the MI300X chip wrote a poem about the city.
Su added, “The more memory you have, the larger the set of models” the device can handle. Customer workloads run faster. We think it’s unique.”
Unlike previous presentations, AMD did not name a key customer for the MI300X or MI300A. The chip’s price and sales boost were not disclosed.
AMD’s shares have doubled this year and hit a 16-month high on Tuesday, but they finished down 3.6% after the AI strategy presentation. Nvidia became the first chipmaker to close above $1 trillion, with a 3.9% gain to $410.22.
I suspect the Street was disappointed by the lack of a significant consumer committing to the MI300 A or X. “They want AMD to say they replaced Nvidia in some design,” said TIRIAS Research chief analyst Kevin Krewell.
Analysts say Nvidia, whose shares have risen 170% this year, leads the AI computing market with 80% to 95%.
Nvidia has few large-scale competitors. Nvidia’s biggest sales threat is Alphabet Inc.’s (GOOGL.O) Google and Amazon.com’s (AMZN.O) cloud unit, which rents their custom chips to outside developers.
AMD claimed it has begun shipping substantial numbers of its “Bergamo” general-purpose central processor chip to companies, including Meta Platforms (META.O).
Facebook parent Meta’s computer infrastructure manager, Alexis Black Bjorlin, said the corporation had embraced the Bergamo chip, which targets a distinct portion of AMD’s data center market that serves cloud computing providers and other large chip customers.
Investors wanted AI news. Nvidia’s edge comes from its chips and more than a decade of offering software tools to AI researchers, anticipating their needs in chips that take years to create. AMD updated its Cuda-competitive Rocm software on Tuesday.
During the presentation, Meta vice president Soumith Chintala said he has worked with AMD to make it easier for AI developers to use free tools to switch from the “single dominating vendor” of AI chips to AMD chips.
“You don’t actually have to do that much work – or almost no work in a lot of cases – to go from one platform to another,” Chintala added.
Analysts warned that just because smart companies like Meta can get good speeds from AMD CPUs doesn’t guarantee market success with less sophisticated clients.
“People still aren’t convinced that AMD’s software solution is competitive with Nvidia’s, even if it’s competitive on the hardware performance side,” said Moor Insights & Strategy analyst Anshel Sag.
